🇬🇧 EN

🔽 Filter

Filters an array to keep only items that match a condition. Use it after a RAG query, HTTP request, or Loop that produces a list, to narrow down results before passing them to another step.

Category: Data Manipulation · Type identifier: filter

Overview

When a workflow receives a large array of items — search results, orders, users, log entries — you often only need a subset of them. The Filter node lets you reduce that array to just the relevant items using a simple condition before passing it on to a Loop, an Agent, or any subsequent step.

You specify a property name on each array item and compare it against a value using one of the built-in operators. The condition is evaluated once per item. Items for which it evaluates to true are kept; all others are discarded. The filtered result and item counts are then available to downstream steps via the standard step output syntax.

Configuration

Field	Status	Description
Array	Required	Reference to the array to filter, e.g. `{{ search.output.results }}`.
Condition Field	Required	The property name on each array item to evaluate, e.g. `score`.
Operator	Required	Comparison operator — `equals`, `not_equals`, `greater_than`, `less_than`, `contains`, `not_contains`, `is_null`, `is_not_null`, `is_empty`, `is_not_empty`.
Value	Optional	The value to compare against. Not needed for `is_null`, `is_not_null`, `is_empty`, `is_not_empty`.

Output Data

The Filter node exposes the following fields on its step output:

Field	Type	Description
`items`	array	Filtered array containing only items that matched the condition.
`count`	number	Number of items in the filtered array.
`excluded_count`	number	Number of items removed by the filter.

Example Usage

Add a RAG node (step key: search) that returns an array of result objects, each with a score property.
Add a Filter node. Set Array to {{ search.output.results }}, Condition Field to score, Operator to greater_than, and Value to 0.8.
Add a Logic Gate after the filter to check {{ filter.output.count }} is greater_than 0 before continuing.
Add an Agent node on the true branch and pass it {{ filter.output.items }} — the agent now processes only high-confidence matches.

Tips & Notes

Filter before you loop. Placing a Filter before a Loop reduces the number of iterations and keeps the loop body simpler than checking conditions inside it.
Empty result is valid. If no items match, items is an empty array and count is 0. Use a Logic Gate after the Filter to handle the empty-result case gracefully.
Null-safe operators. Use is_null or is_not_null to filter items where a field may be absent, avoiding errors that would occur with a direct comparison.
Chaining filters. You can place multiple Filter nodes in sequence to apply several independent conditions — each one narrows the array further.

Related Nodes

Loop — iterate over the filtered array to process each remaining item.
Logic Gate — check the filter count before proceeding, or apply row-level conditions inside a loop.
Set Variable — store or transform data before filtering.
RAG — common source of arrays that benefit from score-based filtering.